Hyperfluid HIP-3 opens a perpetual futures listing to anyone willing to bet $20 million. The question is not whether this democratizes DeFi, but whether the security measures can handle the next developments.
Hyperliquid launched HIP-3 on Mainnet in October 2025, introducing a model in which any builder can deploy perpetual futures markets without committee approval.
Implementers must stake 500,000 HYPE tokens, worth approximately $20 million at current prices, as collateral against malicious behavior.
Validators can lower stakes in whole or in part if a builder charges manipulated prices, recklessly exploits a market, or threatens the network’s solvency. Even during the seven-day non-deployment period, the collateral remains vulnerable to downgrades.
The protocol cuts off HYPE instead of distributing it to users, eliminating incentives for false accusations.
The oracle problem
Builders control the price oracle of their market and fully update the logic, allowing virtually any asset to be brought to market.
Still, it introduces the risk of oracle manipulation, the kind of vulnerability that enabled a $112 million exploit on Mango Markets in 2022, in which an attacker manipulated a thin price feed to drain the platform.
Hyperliquid addresses this by requiring builders to deploy capital large enough to discourage manipulation. The protocol also implements health checks through robust price indices and validator monitoring.
If a market feed fails or a contract expires, builders can invoke a stop function to unwind positions at fair value and freeze trading.
The system assumes that builders will select reliable oracle sources because their deployment depends on it. Validators continuously monitor the markets and can weed out operators that use easily manipulated feeds or allow abnormal operation.
Insulation and insurance
Each market deployed by builders functions as an isolated perpetual exchange with independent order books, margins and risk parameters. Cross-margining with other assets is prohibited to prevent volatility in one market from contaminating other markets.
HIP-3 enforces two types of open interest caps. The first consists of notional ceilings that limit the total dollar value of positions. The second consists of size caps that limit the absolute position size.
These limits apply per asset and worldwide for all assets that a builder lists. Builders can adjust the limits within the limits defined by the protocol, but validators expect conservative defaults for volatile or new assets.
Implementers also set leverage limits and initial margin requirements. The framework prevents a new market from becoming systemically critical overnight.
New markets are launched through a Dutch auction that takes place every 31 hours. Builders bid on HYPE to win deployment slots. To lower barriers to entry, the first three markets a builder enters are exempt from auctions.
Other than winning an auction and wagering 500,000 HYPE, no committee approval is required. Any asset can be listed if the operator has a stake in it. The protocol contains minimum rules for listing.
For example, if a token used as a quote asset for collateral is deemed unsafe, validators can vote to revoke its status, automatically disabling markets that use it.
The steep bond requirement implicitly filters out serious projects with sufficient capital and expertise. Hyperliquid’s documentation states that the goal is to “ensure high-quality markets and protect users” from temporary listings.
Compare approaches
dYdX v4 transitions to permissionless markets, but still requires board votes for new listings. The platform plans to implement isolated margin for risky assets and enforce strict oracle requirements. Assets must be traded on at least six major exchanges to ensure robust price feeds.
Chaos Labs proposed an “asset trial period” with separate insurance funds and closer trading ties for new markets.
GMX v2 addresses similar issues through isolated liquidity pools per trading pair and Chainlink oracles for pricing. The platform integrates Chaos Labs’ Edge Risk Oracle system, which dynamically adjusts open interest limits and price impact coefficients based on real-time conditions.
Furthermore, each GMX market is demarcated, as problems in one group do not affect the other.
Drift Protocol on Solana uses Switchboard’s permissionless oracles to quickly list new assets, but maintains a 10% circuit breaker band.
If the market price deviates more than 10% from the oracle’s five-minute time-weighted average, the market prevents new orders outside that range. Drift also limits individual transactions to a maximum of 2% price impact.
No significant issues were reported during the HIP-3 evaluation phase on testnet. A $21 million theft from Hyperliquid around the same time frame was a private key compromise unrelated to market operations and due to user operational errors.
The real test of the protocol will come when third-party builders deploy new markets for exotic indices or real assets.
Mango Markets collapsed because it allowed a thinly traded token to be used as collateral against a single-source oracle. GMX v1 lost $565,000 when an attacker manipulated AVAX prices off-platform and abused zero-slippage trading.
The design of HIP-3 combines economic deterrence through strikes with technical limitations through caps and insulation. Validators serve as a last resort and can reduce up to 100% of the stake for violations that threaten the correctness or solvency of the network.
The architecture effectively transforms Hyperliquid into a financial infrastructure rather than a single exchange. Each new market functions as its own mini-exchange, secured by the network.
QuickNode’s analysis noted that HIP-3 “replaces gatekeepers with code, while keeping quality and user safety intact through on-chain rules and incentives.”
But who keeps it safe?
The answer is layered. Builders keep the markets safe because their capital is at stake. Validators maintain market safety by providing oversight and reducing authority. The protocol maintains market safety through automated caps, isolation and mental health checks.
This model assumes that rational actors assume that a $20 million bond will deter manipulation more effectively than committee gatekeeping. It assumes that validators will step in when necessary, but the system itself is robust enough that slashing would “never” be necessary on the mainnet, as the Hyperledger team stated.
Lessons learned from Mango and GMX directly formed the basis for these safeguards. Whether the combination of deployment, isolation and supervision can handle all edge cases remains to be proven via live markets.
For now, Hyperliquid offers a simple proposition: Any asset can become a perpetual market if someone believes in it enough to risk $20 million.
The protocol assumes that the price is high enough to distinguish serious builders from reckless ones, and that layered defenses can make up for what economic incentives lack.
